Pymss

Overview

Pymss is a free, open-source toolchain for separating music into vocals, drums, bass, and other stems on your own machine. It suits audio producers who want control over models and processing workflows more than a ready-made online service. Studio brings batch work and stem editing together, but its early development makes it a less reassuring choice for workflows that depend on a stable interface.

Key features

Batch processing and model management

Studio can import files or folders, select a downloaded model, configure inference, and queue jobs. That makes it useful for working through a collection without treating every track as a separate manual task. Its model library can browse, download, resume, and delete model files, and clean up storage; these controls matter when downloaded models begin to accumulate.

Editing and chained workflows

Waveform previews, mix controls, project persistence, and stem export let users review and adjust results in Studio. Pymss can also chain models so that each stage processes the preceding stage's output. This flexibility will appeal to users building multi-step separation workflows, but it is more involved than a service designed around a quick upload and download.

Formats and integrations

The inference package supports WAV, FLAC, MP3, M4A, AAC, Opus, Vorbis, and Ogg output. ComfyUI custom nodes offer another integration route, while the VST3 ARA plugin separates tracks inside an ARA-capable DAW and provides per-stem mute and solo controls. Reaper and Studio One are named as example hosts, so this is a practical fit for users already working in compatible DAWs, not a universal plug-in option.

A browser interface is bundled with the server and shipped with the pip package, and the project also provides API access. The official site says processing runs entirely on the user's machine, which favors local workflows over services centered on remote processing. Support is routed through GitHub documentation and an issue tracker; the Studio README also names a QQ group for Chinese beta testing, bug reports, and discussion.

Pricing

Pymss is free: its plan costs nothing and is open source, with local operation. There is no free trial because the free plan is the product's offering. No paid tier is needed to access the toolchain, so the trade-off is not a reduced quota or a paid upgrade path; users instead take on the setup and workflow choices that come with a local tool.

Licensing varies by repository: Studio is under AGPL-3.0, while the inference package is under MIT. Anyone incorporating or redistributing components should consider the applicable repository license rather than assuming one license covers the whole project.

Platforms

Pymss lists Android, iOS, Linux, macOS, web, Windows, and API support. Its desktop Studio, browser interface, and API provide different ways into the toolchain; the ARA plugin further targets compatible DAWs. Local processing is a central advantage for users who want their audio to stay on their own machine, though the Studio README warns that the app is at an early stage of feature development and breaking changes may occur between versions.

Who it's for

Pymss is best for producers and technically comfortable users who want local stem separation, batch jobs, model management, and options to integrate with a DAW or ComfyUI. It is less suited to someone seeking a polished, stable, low-effort upload service, or to anyone whose setup depends on Studio behavior remaining unchanged across releases.

Pros and cons

• Pro: Local processing keeps separation on the user's machine, a clear fit for users who prefer not to rely on remote processing.
• Pro: Batch queues, model-library controls, project persistence, and stem editing support a fuller production workflow than a one-step split.
• Pro: Multiple output formats, API access, ComfyUI nodes, and an ARA plugin give users several ways to fit separation into existing work.
• Con: Studio is still early in development and may have breaking changes, making it a riskier foundation for workflows that require consistency.
• Con: The ARA plugin depends on an ARA-capable DAW, so it will not serve users working in incompatible hosts.

Alternatives

LALAL.AI is worth considering for a freemium option with a Relaxed Queue offering 10 minutes and a 200 MB per-file upload limit, though its free plan does not allow full downloads. Moises may suit users who prefer a cross-platform freemium service; its free tier caps uploads at five per month, limits files to five minutes, and offers limited stem-separation options.

Fadr is a fit for users who want free MP3 downloads alongside MIDI detection and a remix maker; its Basic plan includes vocals, melodies, drums, and bass. MVSEP is another free choice for users who want access to more than 100 AI models, with the unregistered tier limited to 10 minutes, 100 MB, 50 daily separations, and one concurrent job.

BandLab Splitter suits users looking for browser or mobile access and a free four-stem plan for files under 15 minutes; it also offers a free trial. Sonura is an alternative for users interested in daily credits and limits for loops, one-shots, effects, and full tracks, with free creations released to the public domain under CC0.

StemRoller is another free, open-source option for macOS and Windows that produces vocals, drums, bass, other instrumental stems, and an instrumental track. Suno is aimed at a different choice: its free plan provides 50 credits per day and allows an eight-minute audio upload.

For broader comparisons, see AI Stem Splitters and Stem Separation Software.

Verdict

Choose Pymss if local processing, batch control, and model flexibility matter more to you than a turnkey service. Its free, open-source toolchain combines those strengths with useful editing and integration options. Look elsewhere if Studio's early development and potential breaking changes make a dependable, consistent workflow essential.